J. Chem. In$ Comput. Sei. 1986, 26, 41-52
general speed of the process. There will coexist (with the paper journal) electronic byproducts of the composition system, such as online full text4 and volumes on CD-ROM disk, which will be increasingly available to the more sophisticated computer-supported user. But the overall system will continue to look much as it does today, while the final product-the printed journal-is to all outward appearances identical with that of a half-century ago. Suppose, however, that (1) cost-effective technology becomes widely available for digitizing graphics and half-tones and for printing out completely legible, aesthetically pleasing full pages from digitized files, (2) telecommunication costs drop substantially, because of competition and increased carrying capacity, to the point that it becomes actually less expensive to telecommunicate the content of the manuscript than to mail a paper copy, and (3) coding and compatibility problems are resolved. In such a scenario, conditions would be in place for (1) authors to telecommunicate manuscripts to editors, (2) editors to transmit copy electronically to reviewers, (3) reviewers to send comments on manuscripts to editors by electronic mail, (4) editors to telecommunicate manuscripts to copy editors, and ( 5 ) copy editors to edit online and transmit the manuscript to the printer-with no more inputting necessary. Under these circumstances, the existing system would become both less expensive and quicker, with information moving at electronic speed rather than with the gait of the mailman. At the same time, however, one must recognize that conditions would then exist for a parallel-and conceivably competitive-system, in which authors could post their manuscripts on an electronic bulletin board from which anyone with valid access could read or print out a copy. It seems certain that these technology advances will eventually occur. Some of them have already occurred. But certain
47
questions will need to be answered first before one can confidently predict that an all-electronic primary journal system will (or should) supplant the “mail and paper” system. How will the quality-control function be exercised? How will the archive’s permanence be guaranteed? Will such a system conform to the sociological needs of the scientific author? The author’s personal response to these questions is that he has been for many years a publisher of print journals; as such, it is constitutionally difficult for him to envisage a system that does not center on the print journal. Possibly, the main compulsion to think in this fashion is the difficulty he has conceiving of an electronic archive that has the obvious and imposing permanence of a shelf full of bound volumes of JACS! The main attraction of the present, paper-based system is linked to an author’s unshakeable conviction that his written ideas leave a legacy in ink that the centuries will not erase. This author, for one, does not believe that sheer ease of transmission of ideas will, in practice, override his subconscious fears of leaving his creative legacy in an archive that head crashes or warped disks could render nonexistent in a picosecond! On the other hand, attitudes toward the electronic handling of information are changing so rapidly that any predictions seem, by the very nature of the subject matter, to be speculative and unlikely to be accurate. The next 10 years will, without doubt, provide strong clues to the real answers. REFERENCES AND NOTES See,for example, Shaw, J. G. in Development ofScience Publishing in Europe; Meadows, A. J., Ed.; Elsevier: Amsterdam, 1980; p 149. Bowen, D. H. M. “The Economics of Scientific Journal Publishing”. J. Res. Commun. Stud. 1981, 3, 169-184. Garson, L. R. ‘Computer-Aided Reviewer Selection and Manuscript Control”. Scholarly Publishing 1980, 12, 16-14. Terrant, S. W.; Garson, L. R.; Meyers, B. E.; Cohen, S. “Online Searching: Full Text of American Chemical Society Primary Journals”. J. Chem. Inf Cornput. Sci. 1984, 24, 230-235.
Scientometrics with Some Emphasis on Communication at Scientific Meetings and Through the “Invisible College”+ W. S. LYON Analytical Chemistry Division, Oak Ridge National Laboratory, Oak Ridge, Tennessee 3783 1 Received August 12, 1985 Scientometrics uses quantitative methods to investigate science as an information process. Studies were made of attendance and speakers at several scientific meeting series. Data from these and other investigations lead to the conclusion that “invisible colleges” exist within science and that advancement is often through interactions within these informal organizations. Studies have also been made of what happens to oral presentations (are they eventually published?) and how journals communicate with other journals. Such investigations aid in understanding the communication process in science. Science is an information process, developing with time. As such, it can be investigated quantitatively. The term scientometrics was introduced in 1969 by V. Nalimov’ to stand for those quantitative methods that deal with the investigation of science viewed as an information process. Such studies had been made, of course, for many years before 1969, but the development of computers and computer techniques has made possible rapid search and research that in the past would have taken years to complete. Garfield seized the opportunity ‘Resented before the Division of Chemical Information, 189th National Meeting of the American Chemical Society, Miami Beach, FL,April 19, 1985. Research was sponsored by the Office of Energy Research, US.Department of Energy, under Contract DE-AC05-840R21400 with Martin Marietta Energy Systems, Inc. 0095-233818611626-0047$01.50/0
provided by computer information processing to found IS1 and use of his Citation Indexes and complementary publications is now almost de rigueur for any scientometric study. Terms such as citation rate, impact factor, and immediacy index probably coined and certainly popularized by Garfield have become a part of scientometric jargon even as bit, byte, boot, and Basic have come into the literature through the computer terminal. Whole volumes can be and have been written on scientometric methods; one of the best and most recent is by Braun, and Bujdoso.2 Scientometrics has its own journal called, not suprisingly, Scientometrics, and scientometric papers have appeared in numerous journals including such disparate 0 1986 American Chemical Society
LYON
48 J . Chem. In$ Comput. Sei., Vol. 26, No. 2, 1986 Table I. Probabilities of Attendees Giving Papers at NAA Conferences meeting year(s) probability meeting year(s) probability p67 0.28 p72.18 0.20 p72 0.49 p67.78 0.1 1 Pi a 0.39 p67.72.78 0.056 P,, 0.14 79
publications as Analytical Chemistry, Physics Today, New Scientist, Science, and Nature. No introduction to the subject would be complete without mention of de Solla Price, whose writings include the classic Big Science Little S ~ i e n c eR. , ~ K. Merton, e.g., see reference 4, and Thomas Kuhn whose Structure of Scientifi:c Revolutions5 still has the power to precipitate debate. For a broad discussion of many aspects of scientometric chemistry studies the reader is referred to reference 2. This paper concerns itself with investigations of oral communication at meetings. All work discussed is in the literature, and the interested reader can go to the original paper for more detailed information: ORGANIZATION, ATTENDANCE, SPEAKERS, AND SESSIONS: A STUDY OF FOUR SCIENTIFIC CONFERENCE SERIES Neutron activation analysis (NAA) is an interesting area of science. It is relatively homogeneous, yet in application it crosses almost the entire spectrum of sciences. NAA is a technique that enables one to determine trace element concentrations in materials by irradiation in a nuclear reactor and measurement of radioactive isotopes produced. The International Atomic Energy Agency (IAEA) has sponsored three conferences on Nuclear Activation Techniques in the Life Sciences: 1967, 1972, and 1978. Invitations to participate and paper selection are through formal channels at the national level. Papers are generally limited to about 50, and a quota system is used to assure representation from many countries. Attendance and speakers at these conferences were compared6 to those at three other conference series: IAEA Medical Scintigiaphy (1964, 1968, and 1973), University of Missouri Trace Substances (1969, 1971, and 1976); Air Cleaning Conference (1968, 1972, and 1976). The probability (P,) of a person giving a paper at a meeting is pa = S , / N ,
(1)
where S , = total number of speakers and N , = total number of attendees. The probability of a speaker giving a paper at two conferences is
and at three conferences (3)
Figure 1 shows total and overlapping attendance and speakers at the different IAEA conferences on Nuclear Activation Techniques in the Life Sciences: 1967, 1972, and 1978. The numbers within overlapping arcs indicate multiple-meeting attendees or speakers. Table I shows the probability of attendees giving papers at NAA conferences. If one knows the number of multiple attendees at any two meetings, N,, (as, for example, shown in Figure l ) , one can calculate the expected number of multiple speakers:
sa,
= P,P,N,,
(4)
This calculation has been made for the three dual combinations
Table 11. Calculated and Observed Multiple Speakers at IAEA NAA in Life Sciences“ P , probability N , no. of multiple of giving attendees at multiple speakers vears a Dauer multiple meetings theoretical obsd 67, 72 0.14 39 5 12 44 9 13 72, 78 0.20 61, 78 0.1 1 34 4 8 67, 72, 78 0.056 12 1 3 OTotal speakers, S = 124. Table 111. Calculated and Observed Multiple Speakers at IAEA Medical ScintigraDhv“ P , probability N,no. of multiple of giving attendees at multiple speakers vears a paper multiple meetings theoretical obsd 64, 68 0.116 55 6 26 68, 73 0.066 103 7 30 0.07 41 3 13 64, 73 0.023 33 1 6 64, 68, 73 ‘Total speakers, S = 206. Table IV. Calculated and Observed Multiple Speakers at University of Missouri Trace Substances’ P , probability N,no. of attendees at multiple speakers multiple of giving vears a Daper multiple meetings theoretical obsd 69, 71 0.054 38 2 2 0.052 35 2 I 71, 76 27 2 2 69, 76 0.077 69, 71, 76 0.015 19 0.3 1 “Total speakers, S = 123. Table V. Calculated and Observed Multiple Speakers at Air Cleaning Conferences“ P , probability N,no. of attendees a t multiple speakers multiple of giving years a paDer m u l t i ~ l emeetings theoretical obsd 50 1 6 68, 72 0.017 12, 76 0.037 69 3 9 0.023 33 1 3 68,16 21 0.1 3 68, 72, 76 0.004 Total speakers, S = 140.
shown in Table I1 as well as for the theoretically expected number giving a paper at all three conferences. From Table I1 one sees that the number of observed multiple speakers is almost twice the expected number for two meetings and is 3 times the calculated number for all three. This seems clear evidence that there is a group within NAA that is statistically over represented on programs. This may or may not be desirable. The same calculations were performed for attendees at the IAEA Medical Scintigraphy conferences (Table III), Trace Substances (Table IV), and Air Cleaning Conference (Table V). In every instance, the multiple speakers far exceed the theoretically predicted number. Price3 reintroduced the term “invisible college”, which was first used in the 17th century to refer to a collection of scientists who later formed the Royal Society. “Invisible” suggested that the group members were dispersed and not easily seen. Price used the term to indicate a collection of scientists that through personal accomplishment dominated publications and conferences in a given field. The data of Tables 11-IV tend to support the belief that invisible colleges exist not only in NAA but in other disciplines as well. Other papers reported on institutional and national representation, speakers, and topics at five Modern Trends in Activation Analysis (MTAA) conferences’ and predictions of paper subjects based on data in reference 7 compared with
J . Chem. In$ Comput. Sci., Vol. 26, No. 2, 1986 49
SCIENTOMETRICS
Table VI. Oral Presentation and Publication Data for O R N L Analytical Chemistry Division no. 1st Au no. 1st Au research talks/ no. 1st Au ratio, 1st Au talks total (year Y)/ year, Y talks total talks research total talks papers total 1st Au papers (year Y + 1) 1977 66 37 0.56 36 66/61 = 1.08 1978 60 43 0.72 61 60/71 = 0.85 1979 84 59 0.70 71 84/90 = 0.93 1980 78 63 0.81 90 78/67 = 1.16 1981
92
64
0.70
37/61 43/71 59/90 63/67
Attendance
1972 :99
Y)/
= 0.60 = 0.61 = 0.66 = 0.94
67
= 1.01 h 0.14
1967 = 175
ratio, 1st Au talks research (year 1st Au Paper (year Y + 1)
X = 0.70
* 0.16
Table VII. ACD Research Papers Given and Later Published total talks vear 1977 1978 1979 1980 1981
(N)
37 43 59 63 78
no. published in indicated year, fraction per year Y, Y+1, Y+2, YIN ( U + 1 ) I N ( Y + 2 ) l N 7, 0.19 8, 0.22 8, 0.22 6, 0.14 14, 0.33 0, 0.21 7, 0.12 27, 0.46 3, 0.05 20, 0.32 21, 0.33 3, 0.05
T, TIN 23, 0.62 29, 0.67 37, 0.63 44, 0.70 av: 0.66 f 0.04
VIII. Publishing History of ACD Research Talks Speakers
year presented 1977 1978 1979 1980
1970
:58
Figure 1. Distribution of attendees and speakers a t IAEA NAA meetings.
reality at the 1981 MTAA conference.* Data from these conferences strengthen the belief that invisible colleges do indeed exist. ORAL COMMUNICATION AND INVISIBLE COLLEGES Another way in which oral communications and a small nucleus of colleagues can advance a field was studied in 1981. The field was resonance ionization spectroscopy (RIS), popularly called one-atom detection. In the early stages of development of RIS, oral communications outnumbered printed ones by a ratio of 5 : 1? Total oral presentations hit a peak 4-5 years after invention, and self-citations in the literature sometimes ran as high as 80%. This is not too surprising since RIS was developed and exploited by a small group primarily at one institution (but with a gradually increasing number of outside collaborators). The study does show how a field expands through oral communication and collegial collaboration. A recent paper lo discusses the beginnings of flow injection analysis, about which there has been some controversy. By use of citation data and a scientometric approach, it was postulated that an invisible college associated with a particular group advanced the approach of one claimant as opposed to that of another. Invisible colleges should not be looked upon as sinister forces but rather instead as loose groups of experts who act as leaders and gatekeepers. TECHNICAL ORAL PRESENTATIONS: WHAT HAPPENS TO THEM Technical oral presentations generally fall into the category of research reports or reviews. Research reported is almost
fraction finally published as journal conf proc or article book chapters report 0.22 0.35 0.05 0.28 0.30 0.09 0.27 0.32 0.03 0.42 0.24 0.03
always ongoing or just completed; reviews survey a field or technique, usually include some mention of the speakers' work, and serve as guides or stimulation to other workers. Garvey" and others maintain that these informal channels of communication represent the research front. Obviously for research reports this is true. But what fraction of oral presentations really represent research (as opposed to reviews)? And what is the ultimate fate of such ephemeral efforts? Are they lost, or are they preserved in the archival records of science? Titles of all talks and publications cleared through the O R N L Analytical Chemistry Division were reviewed and categorized as either research or review. The ultimate fate of each presentation was found by examining the publication record of the division for the three succeeding years. A total of 380 oral presentations during the period 1977-1981 were studied.12 Table VI lists total oral presentations, total research talks, and total publications (on a first author basis) for years 1977-1982. A rapid upsurge in all categories was observed in 1979. The ratios of total first author talks (and research first author talks) to first author publications shown in column 6 (and column 7) were obtained by using publication data for the year following the talk. This was felt to be more realistic than using same year data because of the publication time lag between submission of a paper and publication. As seen in Table VI, the ratio of total talks to papers is about one. The ratio of first author research talks to first author papers that resulted from these talks is much lower, 0.70 f 0.16. Thus, it appears that oral presentations are indeed a communication and "trying out" process since about one-third of such efforts never reach the archival literature. Table VI1 traces the publication history of these research talks by year. For example, of the 37 research talks given in 1977, seven were published the same year, eight in 1978, and eight in 1979. The fraction of research papers published is seen to be 0.62 for that year. The picture is fairly consistent with an average of 0.66 f 0.04 talks finally ending as publications.
50 J . Chem. In$ Comput. Sci., Vol. 26, No. 2, 1986
LYON
Table IX. ACS DAC Research Papers Given and Later Published no. published in indicated year, fraction per year WIN) total 1977, 1978. 1979, year talks YIN YlN Y/N TIN 1977 59 7, 0.12 18, 0.30 5 , 0.08 0.51 Table X. Publishing History of ACS DAC Research Talks fraction finally published as year journal conf proc or presented article book chapter report total 1977 0.4 1 0.08 0.02 0.51
Table VI11 shows the fractions of papers published that resulted in journal articles, chapters in conference proceedings or books, and reports. Surprisingly, no consistent pattern of journal and proceedings publication is apparent, and more surprisingly, the fraction appearing in conference proceedings declined drastically in 1980. Results from O R N L were compared with those obtained from an actual conference. The 1977 New Orleans, LA, National American Chemical Society (ACS) meeting and papers given in the Division of Analytical Chemistry (DAC) were selected. Every third paper was taken ( N = 70) and categorized as to research or review, and ultimate publication fate was ascertained by using first author in a computer search of Chemical Abstracts data base. Tables IX and X give results. The ratio of research to review papers are somewhat higher at this meeting (0.84) than that from ORNL with the fraction attaining final publication being somewhat lower (0.51). These data suggest first that research talks are frontier communications and that the failure of all such talks to end up in the literature indicates the health and vigor of the informal communication process and second that published conference proceedings are not usurping the journals’ functions in analytical chemistry. Results of the DAC presentations confirm the hypothesis that such talks are often a trying out process since the ratio research/review is much higher and the fraction of papers published much lower than those from ORNL, which itself requires a somewhat onerous referee and screening process prior to submission of an abstract for conference presentations. Finally, it is rather interesting that from ORNL the yearly ratio of first author research talks to total talks (0.70 f 0.10) is almost identical with the ratio of first author research talks to total papers (0.70 f 0.16). Apparently, the 30% of discussed research that never gets published is balanced by a 30% research output that goes directly from the laboratory to the journal without conference reporting. It is also possible, of course, that some conference-reported research represents rather inconsequential or perhaps dead-end explorations. In any event, an 70% yield does not seem too unsatisfactory even in these days of expanding journals and relatively easy publication. COMMUNICATION AMONG JOURNALS Information obtained from compendia such as Chemical Abstracts and Analytical Abstracts illustrate how fast certain chemical disciplines are growing, which countries are most active in each field, and how many papers journals publish. Use of Citation Index data makes possible additional qualitative and quantitative evaluations. The literature of Analytical C h e m i ~ t r y ,Health ‘~ Physics,I4 and Prompt Nuclear AnalysisI5 were examined. Techniques used were rather simple and straightforward. A more sophisticated approach was used, however, in a study of information flow in analytical
Figure 2. Information flow between analytical chemistry and other fields of science as a percentage of the total flow of analytical information. The areas of the circles representing the fields are proportional to their virtual size.
journals.16 Data from Journal Citation Reports’’ were used to describe information flow (on a macroscale), between two groups of journals and journals in other fields. The two groups of journals were broken down into their components, and the interactions between them described. Finally, the interactions were compared with what might be expected if journals cited themselves as they do others. In this way, one observes how journals communicate with each other. Analytical journals were divided into two groups: (a) 10 broad based (BB) journals, such as Analytical Chemistry and Analyst, and (b) 12 specialty journals, such as International Journal of Mass Spectrometry and Journal of Chromatography. Citations to and from these journals vis-;-vis journals of (a) other chemistry fields and (b) other scientific fields were obtained and information flow diagrams such as that of Figure 2 plotted. The main information sources of analytical chemistry appear to be physics, clinical medicine, and the earth and space sciences; analytical chemistry absorbs information from all these disciplines. By arranging a group of journals in an m X m array, where an element cij indicates both the number of references that journal i gives to journal j and the number of citations that journal j receives from journal i, a reference-citation matrix is obtained. Table XI shows such a matrix for the BB journal. This transaction, or input-output matrix, shows the relationship of a journal to others in the group. For example, the citation ratio for a journal can be a simple indicator of the journal’s behavior within its group. This ratio is found by dividing the sum of the ith column by the sum of the ith row. If the citation ratio is greater than 1, the journal is an emitter or exporter of information; if less than 1, the journal is an absorber or importer of information. For example, within the BB group (Table XI) Analytical Chemistry is seen to be an exporter (215912074 = 1.04); Talanta is an importer (696/836 = 0.83). Such matrices can be the starting point for more sophisticated analyses of information flow. Inspection of the matrix in Table XI reveals an interesting anomaly: most journals cite themselves much more frequently than would be expected. (Self-citations are shown in the diagonal terms.) Such self-citation may well represent a different emphasis, philosophy, or policy from that employed by an author citing articles from other journals. It would be desirable to be able to “normalize” (or correct) this anomalous behavior. Price18 and Price and Burkelg have devised a procedure to replace the diagonal terms by those that would be there if each journal referenced itself as it referenced others. Price’s method consists of treating the matrix as if each ele-
SC~ENTOMETRICS
J. Chem. ZnJ Comput. Sci., Vol. 26, No. 2, 1986 51
Table XI. Cross-Citing of Broad-Based Analytical Chemistry Journals in the Year 1978 citations to ref from
(1) Anal. Chem. (2) Anal. Chim. Acta (3) Analusis (4) Analyst (5) Anal. Lett. (6)Bunseki Kagaku (7)Fresenius' 2.Anal. Chem. (8) Talanta (9)Zavod. Lab. (10)Zh. Anal. Khim. citing sum
1 [3107] 698 103 349 118 200 281 247 27 136 2159
2 578 [541] 37 162 41 74 95 147 11 55 1200
3 46 9 [33]
